Re: [BackupPC-devel] Separating frequently-accessed LOCK and LOG* from $topDir/pc/$host

[Craig B. <cba...@us...>]

Alex,

Thanks for the offer to prepare a pull request, but it doesn't seem worth
making the change.

Have you considered simply reducing your $Conf{WakeupSchedule} so your disk
drive wakes up less often?

I made the DDMMYYYY change in the docs.  Thanks.

Craig

On Tue, Dec 5, 2017 at 1:33 AM, Alexander Kobel <a-...@a-...> wrote:

> Hi,
>
> for a tiny BackupPC instance with on an SSD+HDD-server, I want the HDD to
> be in standby mode (spinned down) most of the time.  Hence, I only store
> __TOPDIR__ on it, and have everything else on the SSD. This almost works,
> except for the fact that unsuccessful pings write a log entry in
> __TOPDIR__/pc/$host/LOG.DDMMYYYY, and LOCK is touched.  Which happens on
> every wakeup when some client is not reachable; in other words, too often.
>
> AFAICS, from what's inside pc/, actually vital for the integrity of the
> backups are the nnn and refCnt dirs (and, to some extent, the backups and
> config.pl files).  Those are only accessed during backup or restore jobs,
> when the "vault" disk is active anyway.
> So, in short, I want to bring __TOPDIR__/pc/$host/LOG* and LOCK from the
> HDD to somewhere else.  Note that both LOG* and nnn are moving targets, so
> symlinks only work to a certain extent.
> (By the way: LOG.DDMMYYYY is actually LOG.MMYYYY, without DD, on my 4.1.3
> installation.)
>
>
> My usual approach is anything-sync-daemon (which uses an overlayfs to move
> some directory to tmpfs, and sync backs periodically and/or on user
> request).  However, this only works at directory level, and would require
> moving the entire __TOPDIR__/pc directory to tmpfs.  nnn can easily become
> large, and overlayfs is ill-suited for that.
> Another approach would be to move __TOPDIR__/pc to the SSD and rsync back
> each __TOPDIR__/pc/$host/ to the HDD in a DumpPostUserCmd and, perhaps,
> after each nightly.  That's better, but again, the nnn can grow large (at
> least in number of files) and might become ill-suited even for the SSD
> and/or rootfs.
>
>
> => Question 1.) Short of patching BackupPC, does anyone have a suggestion
> or alternate solution I didn't think of?
>
>
> It would be way easier to properly separate the two kinds of data if the
> location of the per-host LOG* and LOCK files would be configurable, e.g. to
> somewhere below __LOGDIR__.  This would match the logic for the global LOCK
> (which resides at __LOGDIR__), even though I'd typically expect lock files
> to live below /run/lock or /var/run/lock.
> Full backward compatibility could be kept, e.g., through a
> __PERHOSTLOGDIR__ variable defaulting to __TOPDIR__/pc/.
>
> I'd be willing to look into preparing a patch and pull request, if there
> is any interest.  But I understand that the scenario and motivation are
> negligible in an enterprise setup, and that changing the storage layout can
> a royal pain in the * for existing systems.  Hence,
>
>
> => Question 2.) Would such a change have any chance to be integrated in
> BackupPC?  If so, under what constraints for the interface?  Any
> recommendations for locations?  What about {Xfer,Restore,Archive}{ERR,LOG
> ,Info}*?
>
>
>
> Best,
> Alex
>
>
> ------------------------------------------------------------
> ------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> BackupPC-devel mailing list
> Bac...@li...
> List:    https://lists.sourceforge.net/lists/listinfo/backuppc-devel
> Wiki:    http://backuppc.wiki.sourceforge.net
> Project: http://backuppc.sourceforge.net/
>
>